Early warnings of the timing, style, and size of volcanic eruptions, made on the basis of seismic monitoring, are important for mitigation of the hazards associated with volcanic eruptions. We calculated the rates of seismic energy release at Kelud volcano, Indonesia, preceding both the 2007 effusive lava-dome eruption and the 2014 Plinian eruption. For the 2007 eruption, there was first a step-wise increase in the rate of release of seismic energy, followed by a gradual decrease until the eruption. For the 2014 eruption, the rate of energy release was almost constant until 10 days before the eruption, then gradually increased until there was a dramatic acceleration during the last day before the eruption. The total seismic energy released by the 2007 and 2014 eruptions were 3 x 10(8)J, and 22 x 10(9)J, respectively. The relative difference between the ejected volumes of the 2007 and 2014 eruptions is roughly equivalent to that of the total seismic energy released preceding the eruptions. We conducted several trials of the material failure forecast method (FFM), applying it to the square root of the rate of seismic energy released, to assess its usefulness for predicting the timing of eruptions. The predicted onset of the 2007 eruption was one day earlier than the actual eruption and the seismic energy release preceding the actual eruption did not increase dramatically as suggested by the FFM model. Several trials of FFM for the 2014 eruption predicted the onset of the eruption to be three days to two weeks later than the actual eruption. Even as the end time of the FFM time window approaches several hours to 10 min before the actual eruption onset, the predicted eruption onset is still 3 h later than the actual onset. Thus, the results of our FFM modeling did not show the method to be useful for predicting eruption times for Kelud volcano. The differences of the FFM modeling results for the 2007 and 2014 eruptions likely reflect the differences in the rate of energy release immediately before the two eruptions: a deceleration before the 2007 eruption and an acceleration before the 2014 eruption. These differences may reflect differences in precursory processes for effusive lava-dome and Plinian eruptions. (C) 2017 Elsevier B.V.